🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Ecstatic to host Eldar Kurtic at our ML Theory group (Cohere For AI) next week! Eldar will give us a beginner-friendly introduction to LLM quantization, motivating many of the theoretical aspects behind the techniques used to deliver SoTA systems. 🧵

How does AI memorize information? How does ChatGPT remember your name? How do RNNs keep track of simple numbers? Our new paper led by Barışcan KURTKAYA Fatih Dinc explores the geometry of *short-term memory* in neural networks—revealing limits in how information can be stored!🧵

Join us next week on Friday, March 13th for a presentation by Eldar Kurtic on "Beginner-Friendly Introduction to LLM Quantization: From Zero to Hero." Thanks to Anier Velasco Sotomayor, Thang Chu and Andrej Jovanović for organizing this event 💫

From Yann LeCun on greencard/visa revocations and deportations of scientists.

Reminder 🚨: join us on Thursday evening at 6pm CET to hear Eldar Kurtic talk to us about LLM Quantisation! 🚀

I found it

As a teacher, what else to enjoy more than having two students smarter than yourself?

The more math I learn, the less I want to do anything involving engineering (including machine learning).

A new strongly poly-time algo for negative-weight shortest paths, O~(m sqrt(n)) time, by my colleague Satish Rao. This is the same problem Bellman-Ford, which we teach to undergrads, solved in O(mn) 70 yrs ago. Just 13 pgs! (*puts on todo list to read*) arxiv.org/abs/2503.22613

Join us for a mix of statistical physics and deep learning.

Coming April 10th - new episodes of Google DeepMind: The Podcast! ✨ Join host Professor Hannah Fry as she explores: 🧪 How AI-powered science could revolutionize medicine 🤖 Cutting edge robotics 🗺️ The limitations of human-generated data ↗️ And much more. Subscribe now →

discuss.python.org/t/faster-large… These threads are awesome

Join our ML Theory group next week on April 17th as they welcome Konstantin Schürholt for a session on "Weight Space Learning: Treating Neural Network Weights as Data" Thanks to Anier Velasco Sotomayor, Thang Chu and Andrej Jovanović for organizing this event 🥳

A vector is an element of a vector space. Btw, even in the limited notion of vector that CS people use, a vector ISN’T A SET OF NUMBERS.

We will be presenting this 💫 spotlight 💫 paper at #ICLR2025. Come say hi or DM me if you're interested in discussing AI #interpretability in Singapore! 📆 Poster Session 4 (#530) 🕰️ Fri 25 Apr. 3:00-5:30 PM 📝 openreview.net/forum?id=QogcG… 📊 iclr.cc/virtual/2025/p…

1. GraphAny: Fully-inductive Node Classification on Arbitrary Graphs openreview.net/forum?id=1Qpt4… With Jianan Zhao, Zhaocheng Zhu, Michael Bronstein, and Jian Tang

Long-awaited

We're looking forward to welcoming Jeremy Bernstein for a presentation on "Depths of First-Order Optimization" next week on Thursday, May 8th. Thanks to Anier Velasco Sotomayor Andrej Jovanović and Thang Chu for organizing this session 🤩